MenD crystallizes in an orthorhombic system, space group P
, with four polypeptide chains, labeled A to D in the asymmetric unit. The mass of protein in the tetramer is about 256.4 kDa. The structure was solved by use of single-wavelength anomalous diffraction (SAD) methods and refined to 2.35 Å resolution. The crystallographic statistics and model geometry () indicate that the analysis has produced an acceptable medium-resolution model. Continuous and well-defined electron density was observed for each polypeptide chain from Thr2 to Leu580, except for chain C where the final two residues were not observed. Overlay of the chains using Secondary-Structure Matching20
gives root-mean-square deviations (r.m.s.d.'s) in the range 0.22–0.30 Å, indicating that the subunits are, within the errors associated with a medium-resolution structure, essentially identical. We therefore detail only subunit A, unless stated otherwise.
Crystallographic and model geometry statistics for BsMenD
MenD shows the three-domain architecture typical of ThDP-dependent enzymes ().21
Sequence and structural analyses place MenD in the pyruvate oxidase class of these enzymes, where domain I contains the cofactor pyrimidine-binding motif, domain III binds the diphosphate, and the central domain has variable, often ill-defined, function. In Bs
MenD, domain I spans approximately the first 190 residues. An ordered linker region leads into domain II, which consists of residues 205–345. A smaller linker then leads into domain III, which starts at approximately residue 355 and continues to the C terminus (). Each domain consists of a parallel β-sheet sandwiched between several α-helices; in Bs
MenD, the central domain has five strands and there are six in domains I and III.
Fig. 2 The secondary and domain structure of BsMenD. A ribbon diagram of the enzyme subunit showing the elements of secondary structure with atoms of ThDP depicted as spheres according to atom type: C, gray; N, blue; O, red; S, yellow; P, orange. The N- and (more ...)
Oligomerization is essential for the function of ThDP-dependent enzymes, with each active site formed from two subunits.21
Most of these enzymes function as either dimers or tetramers. Size-exclusion chromatography indicated that Bs
MenD consists of a mixture of dimer and tetramer in solution, with the dimer being much more abundant. The crystal structure of Bs
MenD consists of a dimer-of-dimers (); a similar result was found for Ec
MenD, which is predominantly dimeric in solution15
and then forms a dimer-of-dimers when crystallized.15–17
The BsMenD tetramer. The tetramer is shown as a van der Waals surface with subunits labeled and colored differently. Subunit A is in the same orientation as in .
Analysis using the Protein Interfaces, Surfaces and Assemblies server22
suggests that the dimer-of-dimers is the most stable oligomeric form, burying approximately 23% (5310 Å2
) of the surface area of each subunit (23,100 Å2
). The principal interface, between subunits A and B (and C and D), buries about 14% of the surface area of each subunit. The A:B interface covers an area of about 3130 Å2
, and the C:D interface covers approximately 3160 Å2
. Around 30 hydrogen bonds in addition to the salt bridges formed between Arg53 and Asp55 with their counterparts in the second dimer form the most important interactions in this interface (data not shown). Fewer hydrogen bonds link the dimer-of-dimers arrangement where the pairs of subunits involved are A:D and B:C. This interface is formed primarily by residues located in the linker between domains I and II. A short three-amino-acid stretch of β-strand is present at the start of domain II, which forms an extension of the parallel β-sheet of domain II across the interface. In addition, another three-residue stretch (205–207) forms a two-stranded antiparallel β-sheet across the interface. Approximately 7% of the surface area of a subunit is buried at this interface; 1610 and 1620 Å2
for the A:D and B:C interface regions, respectively. The A:C and B:D interfacial regions occlude about 700 and 710 Å2
, respectively, each representing approximately 2% of the subunit surface area.
Comparison with EcMenD
The MenD proteins from different organisms share little sequence identity, typically in the range 20–30%. Such a low level of sequence conservation despite the same function appears to be a common feature of enzymes that are dependent on ThDP.15
MenD has 28% overall sequence identity to the E. coli
protein. The sequence identity is higher for domains I and III (32% and 33%, respectively) compared to the central domain (16%). Nevertheless, the structures of the two enzymes are closely related, with the majority of secondary-structure elements conserved in each domain (). One exception is the final helix in domain I of Ec
MenD, which is replaced by a loop and β7 in Bs
MenD (). This region does not appear to have any functional role.
Fig. 4 The primary and assigned secondary structure of BsMenD and EcMenD. α-Helices are shown as cylinders and β-strands as arrows, colored according to the domain to which they have been assigned (see ). Residues with a black background (more ...)
Subunit A from Bs
MenD overlays with a subunit from Ec
MenD [Protein Data Bank (PDB) code: 2jlc
] with an r.m.s.d. of approximately 2.0 Å (over 503 aligned Cα
). The closeness of the structures extends to the dimer (r.m.s.d. of 2.0 Å over 1006 aligned Cα
) and the tetramer (2.4 Å over 1996 Cα
). Individual domains align more closely; domain I of Bs
MenD overlays with an r.m.s.d. of 1.3 Å onto domain I of Ec
MenD (173 aligned Cα
), domain II with an r.m.s.d. of 1.8 Å (117 aligned Cα
), and domain III with an r.m.s.d. of 1.7 Å (over 196 aligned Cα
). The loop containing residues responsible for binding to the diphosphate group of ThDP (approximately residues Gly486 to Pro509) is disordered in Ec
MenD in the absence of cofactor16
but ordered in the holo-enzyme.15
The conformation of this cofactor-binding loop in holo-Bs
MenD is similar to that of the corresponding loop in the holo-Ec
structure (data not shown).
The low level of sequence identity between BsMenD and EcMenD serves to highlight the critical determinants of structure and function, and we placed a particular emphasis on conservation when analyzing potential contributions that residues make to enzyme activity. In the following text, unless stated otherwise, we confine our discussion to residues that are strictly conserved between BsMenD and EcMenD.
Cofactor binding and the active site
Dimerization is essential for formation of the cofactor binding and active site, with residues from domain I of subunit A and domain III from subunit B forming one site (and vice versa
, leading to two active sites per dimer, ). A metal cation, assigned as Mn2+
, is bound and helps to position the diphosphate group of ThDP and so tether one part of the cofactor to the protein. This is a common feature of ThDP binding and similar to what is observed in Ec
MenD, the side chains of Asp457 and Asn484, the main-chain carbonyl of Gly486, a water molecule, and two oxygen atoms from the diphosphate group form an octahedral coordination sphere about the metal ion ().
Fig. 5 ThDP interactions with BsMenD. Atoms are colored as follows: C of ThDP, black; C of protein, gray for subunits A and B; P, orange; N, blue; O, red; S, yellow; Mn2+, purple. Purple continuous lines represent coordination of the transition metal ion, and (more ...)
Almost all ThDP-dependent enzymes have a glutamate interacting with N1′ of the pyrimidine; Glu54 fulfills this role in Bs
MenD. Mutagenesis of the corresponding residue, Glu55, in Ec
MenD completely abolished enzyme activity.7
The main-chain amide of Ile433 and carbonyl of Asn431, which participate in hydrogen-bonding interactions with N3′ and N4′, respectively, are from the partner subunit. The side chain of Ile433, a residue strictly conserved in MenD sequences,15
is placed to interact with both the pyrimidine and thiazolium rings helping to hold the cofactor in a V shape, bent at C7′. All ThDP-dependent enzymes that have been structurally characterized have a large hydrophobic side chain in exactly this position. The environment of the cofactor forces N4′ on the pyrimidine ring into close contact with C2 in the thiazolium ring (3.1 Å,
activating the cofactor to contribute to catalysis.
ThDP binding involves contributions from a pair of subunits. The diphosphate is bound between the N-terminal section of α14, strands β16 and β17, and the loop between β17 and α17. The pyrimidine moiety is positioned near the loop linking β15–α15. These are all from one subunit. The partner subunit contributes to binding the pyrimidine with contributions from the C-terminal sections of β1 and β3, the loop between β3 and α4, together with an important contribution from Glu54 at the N terminus of α3. The thiazolium group is sandwiched between Ile433 (on the loop linking β15 with α15) and Phe490 (α17).
The narrow crevice where substrates bind is likewise formed by a dimer. One subunit contributes three α-helical segments (α10, α14, and α17), which line one side of the cavity. Three non-helical segments (the β1–α2 turn, the loop between β6 and β7, and the loop following β4) are provided by the partner subunit. The active site is polar, primarily basic due to the presence of four arginine residues (32A, 106A, 409B, and 428B) and a lysine (299B). A hydrophobic patch is formed by Ile489B, Phe490B, and Leu493B (). Seven of these eight amino acids are strictly conserved, and one, Lys299, is conserved or replaced by arginine in more than 90% of MenD sequences.
Fig. 6 A model for the post-decarboxylation covalent adduct formed after ThDP C2 has reacted with α-ketoglutarate and then isochorismate. A stereoview into the active site with atoms colored as follows: C of ThDP, black; C of protein, gray; O, red; N, (more ...)
Ser405 OG donates a hydrogen bond to one of the phosphate groups of ThDP yet is placed about 3.1 Å from the thiazolium S. This alignment of functional groups may induce a polarization effect on the S atom and contribute to generation of an ylide.
A proposed mechanism
The structure of MenD and our model () of the activated intermediate–isochorismate complex suggested that the actual enzyme mechanism is driven by the chemical properties of the cofactor ThDP. A two-stage mechanism that corresponds to reactions with each of the substrates can be proposed ().15
The absence of any residue that can act as a general acid/base indicates that the cofactor N4′ is a critical component of the mechanism. Such an observation is consistent with structural and mechanistic studies of another ThDP-dependent enzyme, N2
Fig. 7 A two-stage mechanism for catalysis by BsMenD. An asterisk () marks the isochorismate C2, which is attacked by the carbanion intermediate. The post-decarboxylation covalent intermediate in stage II is the structure that has been modeled and is (more ...)
Our model for activity predicts that five polar residues interact with carboxylates and hydroxyl group of substrates to align reagents for catalysis and that non-polar components of isochorismate interact with the hydrophobic patch. The structural studies, sequence comparisons, molecular modeling, and mechanistic considerations lead us to conclude that the environment provided by MenD for the cofactor and the chemical properties of ThDP provide the functionality for catalysis. A first approximation of how substrates bind can then be used to address the molecular recognition of these substrates.
Mutagenesis and kinetic studies
Attempts to soak ligands such as α-ketoglutarate and salicylic acid into the crystals of BsMenD and thus derive structural data directly relevant to substrate recognition resulted in the loss of diffraction. Co-crystallization experiments were also carried out but either no crystals were observed or, where analyzed, there was no electron density that could be attributed to these ligands (data not shown). A model of the post-decarboxylation covalent adduct formed after reaction of ThDP with first α-ketoglutarate then isochorismate ( and ) was constructed, guided by the principle that important features relevant to substrate recognition and catalysis should be conserved between BsMenD and EcMenD. We also noted that the positions of two water molecules and ethane-1,2-diol in the active sites serve to identify potential ligand interacting sites and provide a hint of where functional groups of isochorismate might be placed ().
Modeling of substrate binding, sequence, and structure comparisons identified eight residues that are potentially important with respect to substrate recognition. The first stage of the reaction involves the formation of a covalent adduct between the cofactor and α-ketoglutarate; the formation of such species is well established in ThDP-dependent enzymes24
and the adduct structure has been captured crystallographically.25
The driving force for this stage largely comes from the geometry of the cofactor in the binding site, which brings N4′ of the pyrimidine ring and C2 of the thiazolium ring into close proximity. N4′ abstracts the acidic proton from C2, generating a carbanion ylide, which can then attack the first substrate, α-ketoglutarate, to form the first covalent intermediate (). Arg409 and Arg428 () are well placed to interact with the α-ketoglutaric acid part of this adduct. Mutation of these residues to alanine had a significant effect on the binding of α-ketoglutarate and the rate of reaction (). The Arg409Ala mutation resulted in a 7-fold increase in Km
and 5-fold decrease in kcat
. The Arg428Ala mutation gave an increase in Km
of more than 20-fold and a 4-fold decrease in kcat
. These mutations resulted in comparable kcat
values, which are significantly decreased, compared to the wild-type enzyme. Assay of these two mutants with respect to isochorismate indicates that Arg409 has a greater influence on the second substrate binding since the Km
is increased 8-fold with no change to Km
of the Arg428Ala mutant (). Although catalytic efficiency is compromised in both cases, ascribed to poorer binding of α-ketoglutarate in the first stage of the reaction, the Arg409Ala mutant is affected to a much larger degree with a reduction in kcat
by 2 orders of magnitude. These data suggest then that Arg409 is important for binding of both substrates whereas Arg428 primarily contributes to the binding of α-ketoglutarate. This is consistent with the crystal structure since the side chain of Arg409 is about 1 Å closer to the thiazolium group than Arg428, at the base of the active site cleft directed towards the cavity where the second substrate binds.
Steady-state kinetic parameters with respect to α-ketoglutarate
Steady-state kinetic parameters with respect to isochorismate
Arg32 and Arg106 are in close proximity on one side of the active-site entrance; Lys299 is about 8 Å distant on the other side of the cleft. Arg106 is the closest to the thiazolium group at a distance of about 9 Å. The model of the covalent or activated intermediate suggests that this triumvirate of basic residues is likely to have a small influence, if any, on α-ketoglutarate binding. The kinetic data () indeed suggest that this is the case. There is effectively no change to Km with respect to α-ketoglutarate for the Arg32Ala and Lys299Ala mutants, and Km is reduced by a factor of 3 for the Arg106Ala protein. With respect to isochorismate (), the Lys299Ala mutant displays an increase in Km of about 8-fold but with a comparable kcat value, whereas the mutations to the two arginine residues increase Km between 40- and 50-fold and reduce kcat significantly. The reduction in catalytic efficiency is most pronounced for the Arg106 mutation. These data are consistent with Arg106 performing an important role in recognition and binding of the second substrate.
Three hydrophobic residues, Ile489, Phe490, and Leu493, cluster on one side of the active site (). Of these, Leu493 is most distant, about 9.5 Å from the thiazolium. The kinetic data indicate a minor perturbation of BsMenD activity when Leu493 is mutated to alanine and this is mainly due to an increase in Km for isochorismate. Leu493 does not therefore appear to make a particularly significant or direct contribution to reactivity. The changes to kinetic properties may be due to perturbation of the hydrophobic cluster since the other two residues appear more important. Mutation of Ile489 and Phe490 increases the Km for α-ketoglutarate 16-fold and 12-fold, respectively. The Km for isochorismate is increased by 24- and 25-fold, respectively, and there is a concomitant reduction in catalytic efficiency. Ile489 is adjacent to the site where α-ketoglutarate reacts with the thiazolium group, 3.8 Å distant from the sulfur atom. Phe490 participates in van der Waals interactions with the thiazolium and Ile489. These two residues appear therefore to help bind ThDP. Ile489 is likely to interact with the aliphatic part of the intermediate formed after α-ketoglutarate has reacted with ThDP and, together with Phe490, to interact with the hydrophobic component of the isochorismate ring, C3 and C4 ().
In the second stage of the reaction, the carbanion adduct attacks isochorismate C2 (). Two further arginine residues (Arg32 and Arg106) and a lysine (Lys299) line one side of the putative isochorismate-binding region. A hydrophobic patch consisting of Ile489, Phe490, and Leu493 () lines the opposite side of the active site. Kinetic characterization of Ala mutants of each of these residues is summarized in , with respect to both substrates. Mutation has little effect on the binding of ThDP (data not shown).
Lys299, the only highlighted residue located in the central domain, appears to have little influence on the reaction. Mutation of any of the arginine residues has a similar effect on the binding of isochorismate; however, alteration to Arg106 has a greater impact upon the rate of reaction. This residue is located at the start of the long loop linking β4 with β5; in addition to its putative role in the reaction, Arg106 is well placed to form a hydrogen bond with the main-chain carbonyl of Pro116, helping to orient this loop, which forms part of the binding pocket. Removal of this interaction could destabilize the loop. Of the hydrophobic residues, mutation of Phe490 has the most deleterious effect, with respect to both substrates, indicating particular importance in defining the overall shape and size of the binding pocket.
This combination of single crystal diffraction methods with site-directed mutagenesis and kinetic analyses has allowed us to elucidate the structure–reactivity relationship that governs MenD activity. The enzyme provides the environment to bind the cofactor in a specific manner such that the mechanism is driven by the chemical properties of ThDP. There is no requirement for any amino acid to contribute directly to the mechanism by abstraction or provision of a proton. Rather, the residues in the active site are important for the binding and orientation of substrates allowing MenD to accomplish catalysis. The mutagenesis data are consistent with such a conclusion since they indicate that the most profound effect on catalytic efficiency is derived from alteration of non-polar residues implicated in substrate recognition.